An IGFET is a semiconductor device in which a gate dielectric layer electrically insulates a gate electrode from a channel zone extending between a source zone and a drain zone. The channel zone in an enhancement-mode IGFET is part of a body region, often termed the substrate or substrate region, which forms respective pn junctions with the source and drain. In an enhancement-mode IGFET, the channel zone consists of all the semiconductor material between the source and drain. During IGFET operation, charge carriers move from the source to the drain through a channel induced in the channel zone along the upper semiconductor surface. The threshold voltage is the value of the gate-to-source voltage at which the IGFET starts to conduct current for a given definition of the threshold (minimum) conduction current. The channel length is the distance between the source and drain along the upper semiconductor surface.
IGFETs are employed in integrated circuits (“ICs”) to perform various digital and analog functions. As IC operational capabilities have advanced over the years, IGFETs have become progressively smaller, leading to a progressive decrease in minimum channel length. An IGFET that operates in the way prescribed by the classical model for an IGFET is often characterized as a “long-channel” device. An IGFET is described as a “short-channel” device when the channel length is reduced to such an extent that the IGFET's behavior deviates significantly from the classical IGFET model. Although both short-channel and long-channel IGFETs are employed in ICs, the great majority of ICs utilized for digital functions in very large scale integration applications are laid out to have the smallest channel length reliably producible with available lithographic technology.
A depletion region extends along the junction between the source and the body region. Another depletion region extends along the junction between the drain and the body region. A high electric field is present in each depletion region. Under certain conditions, especially when the channel length is small, the drain depletion region can laterally extend to the source depletion region and merge with it along or below the upper semiconductor surface. The merging of the source and drain depletion regions along the upper semiconductor surface is termed surface punchthrough. The merging of the two depletion regions below the upper semiconductor surface is termed bulk punchthrough. When surface or bulk punchthrough occurs, the operation of the IGFET cannot be controlled with its gate electrode. Both types of punchthrough need to be avoided.
Various techniques have been employed to improve the performance of IGFETs, including those operating in the short-channel regime, as IGFET dimensions have decreased. One performance improvement technique involves providing an IGFET with a two-part drain for reducing the electric field at the drain so as to avoid hot carrier injection into the gate dielectric layer. The IGFET is also commonly provided with a similarly configured two-part source. Another conventional performance improvement technique is to increase the dopant concentration of the channel zone in a pocket portion along the source for inhibiting surface punchthrough as channel length is reduced and for shifting generally undesired roll-off of the threshold voltage to shorter channel length. Similar to how the IGFET is provided with a two-part source analogous to the two-part drain, the dopant concentration is also commonly increased in a pocket portion along the drain. The resulting IGFET is then typically a symmetric device.
FIG. 1 illustrates such a conventional long-channel symmetric n-channel IGFET 20 as described in U.S. Pat. No. 6,548,842 B1 (Bulucea et al.). IGFET 20 is created from a p-type monocrystalline silicon (“monosilicon”) semiconductor body. The upper surface of IGFET 20 is provided with recessed electrically insulating field-insulating region 22 that laterally surrounds active semiconductor island 24 having n-type source/drain (“S/D”) zones 26 and 28. Each S/D zone 26 or 28 consists of very heavily doped main portion 26M or 28M and more lightly doped, but still heavily doped, lateral extension 26E or 28E.
S/D zones 26 and 28 are separated from each other by channel zone 30 of p-type body material 32 consisting of lightly doped lower portion 34, heavily doped intermediate well portion 36, and upper portion 38. Although most of upper body-material portion 38 is moderately doped, portion 38 includes ion-implanted heavily doped halo pocket portions 40 and 42 that respectively extend along S/D zones 26 and 28. IGFET 20 further includes gate dielectric layer 44, overlying very heavily doped n-type polycrystalline silicon (“polysilicon”) gate electrode 46, electrically insulating gate sidewall spacers 48 and 50, and metal silicide layers 52, 54, and 56.
S/D zones 26 and 28 are largely mirror images of each other. Halo pockets 40 and 42 are also largely mirror images of each other so that channel zone 30 is symmetrically longitudinally graded with respect to channel dopant concentration. Due to the symmetry, either S/D zone 26 or 28 can act as source during IGFET operation while the other S/D zone 28 or 26 acts as drain. This is especially suitable for some digital situations where S/D zones 26 and 28 respectively function as source and drain during certain time periods and respectively as drain and source during other time periods.
FIG. 2 illustrates how net dopant concentration NN varies as a function of longitudinal distance x for IGFET 20. Since IGFET 20 is a symmetric device, FIG. 2 presents only a half profile starting from the channel center. Curve segments 26M*, 26E*, 28M*, 28E*, 30*, 40*, and 42* in FIG. 2 respectively represent the net dopant concentrations of regions 26M, 26E, 28M, 28E, 30, 40, and 42. Dotted curve segment 40″ or 42″ indicates the total concentration of the p-type semiconductor dopant that forms halo pocket 40 or 42, including the p-type dopant introduced into the location for S/D zone 26 or 28 in the course of forming pocket 40 or 42.
The increased p-type dopant channel dopant concentration provided by each halo pocket 40 or 42 along S/D zone 26 or 28, specifically along lateral S/D extension 26E or 28E, causes surface punchthrough to be avoided. Upper body-material portion 38 is also provided with ion-implanted p-type anti-punchthrough (“APT”) semiconductor dopant that reaches a maximum concentration in the vicinity of the depth of S/D zones 26 and 28. This causes bulk punchthrough to be avoided.
Based on the information presented in U.S. Pat. No. 6,548,842, FIG. 3a roughly depicts how concentrations NT of the total p-type and total n-type dopants vary as a function of depth y along an imaginary vertical line extending through main S/D portion 26M or 28M. Curve segment 26M″ or 28M″ in FIG. 3a represent the total concentration of the n-type dopant that defines main S/D portion 26M or 28M. Curve segments 34″, 36″, 38″, 40″, and 42 together represent the total concentration of the p-type dopant that defines respective regions 34, 36, 38, 40, and 42.
Well portion 36 is defined by ion implanting IGFET 20 with p-type main well semiconductor dopant that reaches a maximum concentration at a depth below that of the maximum concentration of the p-type APT dopant. Although, the maximum concentration of the p-type main well dopant is somewhat greater than the maximum concentration of the p-type APT dopant, the vertical profile of the total p-type dopant is relatively flat from the location of the maximum well-dopant concentration up to main S/D portion 26M or 28M. U.S. Pat. No. 6,548,842 discloses that the p-type dopant profile along the above-mentioned vertical line through main S/D portion 26M or 28M can be further flattened by implanting an additional p-type semiconductor dopant that reaches a maximum concentration at a depth between the depths of the maximum concentrations of APT and well dopants. This situation is illustrated in FIG. 3b where curve segment 58″ indicates the variation caused by the further p-type dopant.
Body material 32 is alternatively referred to as a well because it is created by introducing p-type semiconductor dopant into lightly doped semiconductor material of a semiconductor body. The so-introduced total well dopant here consists of the p-type main well dopant, the APT dopant, and, in the IGFET variation of FIG. 3b, the additional p-type dopant.
Various types of wells have been employed in ICs, particularly ICs containing complementary IGFETs where wells must be used for either the n-channel or p-channel IGFETs depending on whether the lightly doped starting semiconductor material for the IGFET body material is of p-type or n-type conductivity. ICs containing complementary IGFETs commonly use both p-type and n-type wells in order to facilitate matching of n-channel and p-channel IGFET characteristics.
Early complementary-IGFET (“CIGFET”) fabrication processes commonly termed “CMOS” fabrication often created wells, referred to here as “diffused” wells, by first introducing main semiconductor well dopant shallowly into lightly doped semiconductor material prior to formation of a recessed field-insulating region typically consisting largely of thermally grown silicon oxide. Because the field-oxide growth was invariably performed at high temperature over a multi-hour period, the well dopant diffused deeply into the semiconductor material. As a result, the maximum concentration of the diffused well dopant occurred at, or very close to, the upper semiconductor surface. Also, the vertical profile of the diffused well dopant was relatively flat near the upper semiconductor surface.
In more recent CIGFET fabrication processes, ion implantation at relatively high implantation energies has been utilized to create wells subsequent to formation of the field oxide. Since the well dopant is not subjected to the long high-temperature operation used to form the field oxide, the maximum concentration of the well dopant occurs at a significant depth into the semiconductor material. Such a well is referred to as a “retrograde” well because the concentration of the well dopant decreases in moving from the subsurface location of the maximum well-dopant concentration to the upper semiconductor surface. Retrograde wells are typically shallower than diffused wells. The advantages and disadvantages of retrograde wells are discussed in (a) Brown et al., “Trends in Advanced Process Technology—Submicrometer CMOS Device Design and Process Requirements”, Procs. IEEE, December 1986, pp. 1678-1702, and (b) Thompson et al., “MOS Scaling: Transistor Challenges for the 21st Century”, Intel Technology J, Q398, 1998, pp. 1-19.
FIG. 4 illustrates symmetric n-channel IGFET 60 that employs a retrograde well as generally described in Rung et al. (“Rung”), “A Retrograde p-Well for Higher Density CMOS”, IEEE Trans Elec. Devs., October 1981, pp. 1115-1119. Regions in FIG. 4 corresponding to regions in FIG. 1 are, for simplicity, identified with the same reference symbols. With this in mind, IGFET 60 is created from lightly doped n-type substrate 62. Recessed field-insulating region 22 is formed along the upper semiconductor surface according to the local-oxidation-of-silicon process. P-type retrograde well 64 is subsequently formed by selectively implanting p-type semiconductor dopant into part of substrate 62. The remaining IGFET regions are then formed to produce IGFET 60 as shown in FIG. 4.
The p-type dopant concentration of retrograde well 64 is at moderate level, indicated by the symbol “p”, in the vicinity of the peak well dopant concentration. The well dopant concentration drops to a low level, indicated by the symbol “p−” at the upper semiconductor surface. The dotted line in FIG. 4 indicates generally where the well dopant concentrations transitions from the p level to the p− level in moving from the p portion of well 64 to the upper semiconductor surface.
FIG. 5 indicates the general nature of the dopant profile along an imaginary vertical line through the longitudinal center of IGFET 60 in terms of net dopant concentration NN. Curve segments 62″ and 64″ respectively represent the net dopant concentrations of n-type substrate 62 and p-type retrograde well 64. Arrow 66 indicates the location of the maximum subsurface p-type dopant concentration in well 64. For comparison, curve segment 68″ represents the vertical dopant profile of a typical deeper p-type diffused well.
A specific example of the dopant profile along an imaginary vertical line through the longitudinal center of retrograde well 64 as simulated by Rung is depicted in FIG. 6 in terms of net dopant concentration NN. Curve segment 26″ or 28″ indicates the net dopant concentration along an imaginary vertical line through S/D zone 26 or 28 of Rung's simulation of IGFET 60. As FIG. 6 indicates, the concentration of the p-type well dopant decreases by more than a factor of 10 in moving from location 66 of the maximum p-type dopant concentration in well 64 to the upper semiconductor surface. FIG. 6 also indicates that the depth of location 66 is approximately twice as deep as S/D zone 26 or 28 in IGFET 60.
A retrograde IGFET well, such as well 64, whose maximum well dopant concentration (i) is at least a factor of 10 greater than the well dopant concentration at the upper semiconductor surface and (ii) occurs relatively deep compared to, e.g., deeper than, the maximum depth of the S/D zones can be viewed as an “empty” well since there is a relatively small amount of well dopant near the top of the well where the IGFET's channel forms. In contrast, a diffused well is a “filled” well. The well for symmetric IGFET 20 in FIG. 1 can likewise be viewed as a filled well since the APT dopant “fills” the retrograde well that would otherwise occur if the main well dopant were the only well dopant.
A symmetric IGFET structure is generally not needed in situations where current flows in only one direction through an IGFET during device operation. As further discussed in U.S. Pat. No. 6,548,842, drain-side halo pocket portion 42 of symmetric IGFET 20 can be deleted to produce long n-channel IGFET 70 as shown in FIG. 7a. IGFET 70 is an asymmetric device because channel zone 30 is asymmetrically longitudinally dopant graded. S/D zones 26 and 28 in IGFET 70 respectively function as source and drain. FIG. 7b illustrates asymmetric short n-channel IGFET 72 corresponding to long-channel IGFET 70. In IGFET 72, source-side halo pocket 40 closely approaches drain 28. Net dopant concentration NN as a function of longitudinal distance x along the upper semiconductor surface is shown in FIGS. 8a and 8b respectively for IGFETs 70 and 72.
Asymmetric IGFETs 70 and 72 receive the same APT and well implants as symmetric IGFET 60. Along vertical lines extending through source 26 and drain 28, IGFETs 70 and 72 thus have the dopant distributions shown in FIG. 3a except that dashed-line curve segment 74″ represents the vertical dopant distribution through drain 28 due to the absence of halo pocket 42. When the IGFET structure is provided with the additional well implant to further flatten the vertical dopant profile, FIG. 3b presents the consequent vertical dopant distributions again subject to curve segment 74″ representing the dopant distribution through drain 28.
U.S. Pat. Nos. 6,078,082 and 6,127,700 (both Bulucea) describe IGFETs having asymmetric channel zones but different vertical dopant characteristics than those employed in the inventive IGFETs of U.S. Pat. No. 6,548,842. IGFETs having asymmetric channel zones are also examined in other prior art documents such as (a) Buti et al., “Asymmetrical Halo Source GOLD drain (HS-GOLD) Deep Sub-half n-Micron MOSFET Design for Reliability and Performance”, IEDM Tech. Dig., 3-6 Dec. 1989, pp. 26.2.1-26.2.4, (b) Chai et al., “A Cost-Effective 0.25 μm Leff BiCMOS Technology Featuring Graded-Channel CMOS (GCMOS) and a Quasi-Self Aligned (QSA) NPN for RF Wireless Applications”, Procs. 2000 Bipolar/BiCMOS Circs. and Tech. Meeting, 24-26 Sep. 2000, pp. 110-113, (c) Ma et al., “Graded-Channel MOSFET (GCMOSFET) for High Performance, Low Voltage DSP Applications”, IEEE Trans. VLSI Systs. Dig, December 1997, pp. 352-358, (d) Su et al., “A High-Performance Scalable Submicron MOSFET for Mixed Analog/Digital Applications”, IEDM Tech. Dig., December 1991, pp. 367-370, and (e) Tsui et al., “A Volatile Half-Micron Complementary BiCMOS Technology for Microprocessor-Based Smart Power Applications”, IEEE Trans. Elec. Devs., March 1995, pp. 564-570.
Choi et al. (“Choi”), “Design and analysis of a new self-aligned asymmetric structure for deep sub-micrometer MOSFET”, Solid-State Electronics, Vol. 45, 2001, pp. 1673-1678, describes an asymmetric n-channel IGFET configured similarly to IGFET 70 or 72 except that the source extension is more heavily doped than the drain extension. Choi's IGFET also lacks a well region corresponding to intermediate well portion 36. FIG. 9 illustrates Choi's IGFET 80 using the same reference symbols as used for IGFET 70 or 72 to identify corresponding regions. Although source extension 26E and drain extension 28E are both labeled “n+” in FIG. 9, the doping in source extension 26E of IGFET 80 is somewhat more than a factor of 10 greater than the doping in drain extension 28E. Choi indicates that the heavier source-extension doping should reduce the increased source-associated parasitic capacitance that otherwise results from the presence of halo pocket 40 along source 26.
FIGS. 10a-10d (collectively “FIG. 10”) represent steps in Choi's process for fabricating IGFET 80. Referring to FIG. 10a, precursors 44P and 46P respectively to gate-dielectric layer 44 and polysilicon gate electrode 46P are successively formed along p-type monosilicon wafer 34P that constitutes a precursor to body-material portion 34. A layer of pad oxide is deposited on precursor gate-electrode layer 46P and patterned to produce pad oxide layer 82. A layer of silicon nitride is deposited on top of the structure and partially removed to produce nitride region 84 that laterally abuts pad oxide 82 and leaves part of gate-electrode layer 46P exposed.
After removing the exposed part of gate-electrode layer 46P, singly ionized arsenic is ion implanted through the exposed part of dielectric layer 44P and into wafer 34P at an energy of 10 kiloelectron volts (“keV”) and a high dosage of 1×1015 ions/cm2 to define heavily doped n-type precursor 26EP to source extension 26E. See FIG. 10b. Singly ionized boron difluoride is also ion implanted through the exposed part of dielectric layer 44P and into wafer 34P to define heavily doped p-type precursor 40P to source-side halo pocket 40. The halo implantation is done at an energy of 65 keV and a high dosage of 2×1013 ions/cm2.
Nitride region 84 is converted into silicon nitride region 86 that laterally abuts pad oxide 82 and covers the previously exposed part of dielectric layer 44P. See FIG. 10c. After removing pad oxide 82, the exposed part of gate-electrode layer 46P is removed to leave the remainder of layer 46P in the shape of gate electrode 46 as shown in FIG. 10d. Another part of dielectric layer 44P is thereby exposed. Singly ionized arsenic is ion implanted through the newly exposed part of dielectric layer 44P and into wafer 34P to define heavily doped n-type precursor 28EP to drain extension 28E. The drain-extension implant is done at the same energy, 10 keV, as the source extension implant, but at a considerably lower dosage, 5×1013 ions/cm2. As a result, the drain-extension and source-extension implants reach maximum concentrations at essentially the same depth into wafer 34P. In later steps (not shown), nitride 86 is removed, gate sidewall spacers 48 and 50 are formed, arsenic is ion implanted to define n++ main S/D portions 26M and 28M, and a rapid thermal anneal is performed to produce IGFET 80 as shown in FIG. 9.
Choi's decoupling of the source-extension and drain-extension implants and then forming source extension 26E at a considerably higher doping than drain extension 28E in order to alleviate the increased source-associated parasitic capacitance resulting from source-side halo pocket 40 is clearly advantageous. However, Choi's coupling of the formation of gate electrode 46 with the formation of source/drain extensions 26E and 28E in the process of FIG. 10 is laborious and could make it difficult to incorporate Choi's process into a larger semiconductor process that provides other types of IGFETs. It would be desirable to have a simpler technique for making such an asymmetric IGFET. In particular, it would be desirable to decouple the gate-electrode formation from the formation of differently doped source/drain extensions.
The term “mixed signal” refers to ICs containing both digital and analog circuitry blocks. The digital circuitry typically employs the most aggressively scaled n-channel and p-channel IGFETs for obtaining the maximum potential digital speed at given current leakage specifications. The analog circuitry utilizes IGFETs and/or bipolar transistors subjected to different performance requirements than the digital IGFETs. Requirements for the analog IGFETs commonly include high linear voltage gain, good small-signal and large-signal frequency response at high frequency, good parameter matching, low input noise, well controlled electrical parameters for active and passive components, and reduced parasitics, especially reduced parasitic capacitances. Although it would be economically attractive to utilize the same transistors for the analog and digital blocks, doing so would typically lead to weakened analog performance. Many requirements imposed on analog IGFET performance conflict with the results of digital scaling.
More particularly, the electrical parameters of analog IGFETs are subjected to more rigorous specifications than the IGFETs in digital blocks. In an analog IGFET used as an amplifier, the output resistance of the IGFET needs to be maximized in order to maximize its intrinsic gain. The output resistance is also important in setting the high-frequency performance of an analog IGFET. In contrast, the output resistance is considerably less importance in digital circuitry. Reduced values of output resistance in digital circuitry can be tolerated in exchange for higher current drive and consequent higher digital switching speed as long as the digital circuitry can distinguish its logic states, e.g., logical “0” and logical “1”.
The shapes of the electrical signals passing through analog transistors are critical to circuit performance and normally have to be maintained as free of harmonic distortions and noise as reasonably possible. Harmonic distortions are caused primarily by non-linearity of transistor gain and transistor capacitances. Hence, linearity demands on analog transistors are very high. The parasitic capacitances at pn junctions have inherent voltage non-linearities that need to be alleviated in analog blocks. Conversely, signal linearity is normally of secondary importance in digital circuitry.
The small-signal analog speed performance of IGFETs used in analog amplifiers is determined at the small-signal frequency limit and involves the small-signal gain and the parasitic capacitances along the pn junctions for the source and drain. The large-signal analog speed performance of analog amplifier IGFETS is similarly determined at the large-signal frequency limit and involves the non-linearities of the IGFET characteristics.
The digital speed of logic gates is defined in terms of the large-signal switching time of the transistor/load combination, thereby involving the drive current and output capacitance. Hence, analog speed performance is determined differently than digital speed performance. Optimizations for analog and digital speeds can be different, leading to different transistor parameter requirements.
Digital circuitry blocks predominantly use the smallest IGFETs that can be fabricated. Because the resultant dimensional spreads are inherently large, parameter matching in digital circuitry is often relatively poor. In contrast, good parameter matching is usually needed in analog circuitry to achieve the requisite performance. This typically requires that analog transistors be fabricated at greater dimensions than digital IGFETs subject to making analog IGFETS as short as possible in order to have source-to-drain propagation delay as low as possible.
In view of the preceding considerations, it is desirable to have a semiconductor fabrication platform that provides IGFETs with good analog characteristics. The analog IGFETs should have high intrinsic gain, high output resistance, high small-signal switching speed with reduced parasitic capacitances, especially reduced parasitic capacitances along the source-body and drain-body junctions. It is also desirable that the fabrication platform be capable of providing high-performance digital IGFETs.